Investigation of DCT/PCA combined with Kohonen classifier for human identification

In this work, we use DCT transform and eigen-faces method to describe our facial dataset and to build a face recognition system using Kohonen self organizing map neural network classifier. Ascendant hierarchical classification is used to describe the efficiency of DCT transform. A Kohonen SOM neural network classifier is used to classify an individual from three face databases: The ORL, computer vision and Caltech databases. These data sets are used to illustrate the effect of variation of illumination, background and facial rotation. Our study demonstrates the efficiency of DCT without PCA reduction especially in controlled environment.